Skip to content

Latest commit

 

History

History
134 lines (101 loc) · 7.42 KB

CONTRIBUTING.md

File metadata and controls

134 lines (101 loc) · 7.42 KB

Why would I want to contribute?

This project is intended to be developed and maintained by a commercial company as a component with no goal of becoming commercial in itself. It is more of a public utility that we hope will be useful to many people and companies.

I’m in! Now what?

Join the Discord Server!.

Principles

  • Less prompt is more. The API user should have as much control over the prompts and we should hard code as little as possible.
  • Follow strictly OpenAI API. We should rarely deviate or add to the API. We should only do so when it is clear that the API is lacking.
  • Rely on open source projects that survived the test of time.
  • Rely on cloud native technologies that survived the test of time.
  • Standalone and edge ready. The API should be able to run on a Raspberry Pi for example.
  • Test as much as possible. We should have a high test coverage.
  • Measure. We should maintain benchmarks of performance and cost over different variables (models, hardware, tasks, etc.)

Current Stack

Infra

  • Postgres for all the cold data (retrieval, functions, assistants, etc.)
  • Redis for all the hot data (atm only queuing runs, next could be used for caching)
  • Minio for all the files
  • Docker

Backend

  • Rust, Axum, etc.

Taking on Tasks

We have a growing task list of issues. Find an issue that appeals to you and make a comment that you'd like to work on it. Include in your comment a brief description of how you'll solve the problem and if there are any open questions you want to discuss.

If the issue is currently unclear but you are interested, please post in Discord and someone can help clarify the issue in more detail.

Setup

  1. Install Rust curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
  2. Install Docker
  3. Run docker services postgres, redis, and minio docker compose -f docker/docker-compose.yml up postgres redis minio
  4. Run the server and runs executor: make all (it's running cargo under the hood which is the package manager of Rust)

"runs executor" is just a piece of code that wait items being added in Redis and execute runs when something is added.

We recommend using Cursor for programming - best IDE nowadays. Also make sure to install Rust extension.

Submitting Work

We're all working on different parts of HAL-9100 together. To make contributions smoothly we recommend the following:

  1. Fork this project repository and clone it to your local machine. (Read more About Forks) or use gitpod.io.
  2. Before working on any changes, try to sync the forked repository to keep it up-to-date with the upstream repository.
  3. On a new branch in your fork (aka a "feature branch" and not main) work on a small focused change that only touches on a few files.
  4. Package up a small bit of work that solves part of the problem into a Pull Request and send it out for review.
  5. If you're lucky, we can merge your change into main without any problems. If there are changes to files you're working on, resolve them by:
    1. First try to rebase as suggested in these instructions.
    2. If rebasing feels too painful, merge as suggested in these instructions.
  6. Once you've resolved conflicts (if any), finish the review and squash and merge your PR (when squashing try to clean up or update the individual commit messages to be one sensible single one).
  7. Merge in your change and move on to a new issue or the second step of your current issue.

Additionally, if someone is working on an issue that interests you, ask if they need help on it or would like suggestions on how to approach the issue. If so, share wildly. If they seem to have a good handle on it, let them work on their solution until a challenge comes up.

Releasing new Docker images

Just commit with Release and version number in the commit message. For example:

git commit -m "Release 1.0.0"

Once the image is built, preferrably update the Docker compose file to use the new version.

Tips

  • At any point you can compare your feature branch to the upstream/main of llm-edge/hal-9100 by using a URL like this: https://github.com/llm-edge/hal-9100/compare/main...bobm4894:hal-9100:my-example-feature-branch. Obviously just replace bobm4894 with your own GitHub user name and my-example-feature-branch with whatever you called the feature branch you are working on, so something like https://github.com/llm-edge/hal-9100/compare/main...<your_github_username>:hal-9100:<your_branch_name>. This will show the changes that would appear in a PR, so you can check this to make sure only the files you have changed or added will be part of the PR.
  • Try not to work on the main branch in your fork - ideally you can keep this as just an updated copy of main from llm-edge/hal-9100.
  • If your feature branch gets messed up, just update the main branch in your fork and create a fresh new clean "feature branch" where you can add your changes one by one in separate commits or all as a single commit.
  • When working on Github actions, you can test locally using act like so act -W .github/workflows/ci_core.yml --container-architecture linux/amd64 (container-architecture is necessary if you use Mac M series)

When does a review finish

A review finishes when all blocking comments are addressed and at least one owning reviewer has approved the PR. Be sure to acknowledge any non-blocking comments either by making the requested change, explaining why it's not being addressed now, or filing an issue to handle it later.

Troubleshooting

  • Sometimes you need to run cargo sqlx prepare --workspace - typically upon writing sql queries or editing the migrations. SQLX runs checks at build time, welcome to Rust!
  • make check helps you check what's in the db, queue, files, etc.
  • make reboot helps you reset the db, queue, files, etc.
  • use test-driven development
  • use breakpoints and debuggers to see step-by-step what's happening and what's not happening
  • DM @louis030195 on any channel to get help